1 / 54

Building Blocks for a Simple TeraGrid Science Gateway

Building Blocks for a Simple TeraGrid Science Gateway. Yan Liu Shaowen Wang CyberInfrastructure and Geospatial Information Laboratory (CIGI) ‏ and National Center for Supercomputing Applications (NCSA) ‏ University of Illinois at Urbana-Champaign SciDAC’09, June 19, 2009, San Diego, CA.

tommy
Download Presentation

Building Blocks for a Simple TeraGrid Science Gateway

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Building Blocks for a Simple TeraGrid Science Gateway Yan Liu Shaowen Wang CyberInfrastructure and Geospatial Information Laboratory (CIGI)‏ and National Center for Supercomputing Applications (NCSA)‏ University of Illinois at Urbana-Champaign SciDAC’09, June 19, 2009, San Diego, CA

  2. Objectives • Create a user-centric Web 2.0 environment as science gateway interface • Use TeraGrid to support domain-specific scientific computing • Develop Grid-enabled application services to access TeraGrid capabilities • Understand a GISolve-based workflow to steer analyses on TeraGrid http://gisolve.org/ http://www.cigi.uiuc.edu/doku.php/projects/simplegrid/index

  3. SimpleGrid Overview Gateway User Environment Web Services Web 2.0 User Interface Collaborative Analysis Visualization Development Sharing Automatic Grid-enabled Applications SimpleCred SimpleRun SimpleTran SimpleViz Automatic Manual TeraGrid AAAA Computation Storage Information Learning

  4. Learning Curve • Case study: a Grid-based GIScience computing analysis application (DMS)‏ • Simplified from a production GISolve application • Three-stage learning • Building a simple &user-friendly science gateway interface using Web 2.0 • Accessing TeraGrid resources • Streamlining Grid-based scientific computing as Grid-enabled Web services

  5. Progressive Learning • Build a simple science gateway user environment • User-friendly Web 2.0 interface for user interactions • Collaborative analysis using Twitter • Google Map-based visualization • Integrated access for analysis, collaboration, and visualization • Access TeraGrid resources • Accounts and allocations • Computing • Data storage • Information services • Develop Grid-enabled applications • Globus Toolkit programming on TeraGrid access • Sharing Grid-enabled applications as Web services

  6. Development Curve • Science gateway • Yahoo UI (YUI)-based Web 2.0 user environment • JavaScript programming on user interface components • Ajax communications • Php-based server code programming • A straightforward GIScience (Geographic Information Science) collaborative application workflow • TeraGrid middleware and services for domain-specific computation • Deploying and running domain science applications on TeraGrid • Grid-enabled application • Globus Toolkit programming for TeraGrid access • Application-specific SimpleGrid API • Axis2-based Web service development • Reusable components for development • SimpleGrid APIs • Designed to be SIMPLE • Manuals for SimpleGrid setup • Useful links for further study

  7. Application Example • Application: the DMS Spatial Analysis • Spatial interpolation based on the Dynamically Memorized Strategy (DMS) • Application package • Binary executables • Sample datasets • Build user-friendly science gateway interfaces for scientists to perform DMS analysis transparently on TeraGrid • Developing user-friendly gateway Web interface to meet user interaction requirements • Using TeraGrid middleware and services for analysis computation on TeraGrid resources • Streamlining the computation process by developing the Grid-enabled DMS application • Sharing DMS application by relevant science communities as a Web service and integrating the service into gateway

  8. Outline • Part I: SimpleGrid gateway demonstration • Part II: SimpleGrid user environment development • Part III: TeraGrid services and tools • Part IV: SimpleGrid API and Web service development for Grid-enabled analysis

  9. Part I: SimpleGrid Gateway Demonstration • Science gateways to TeraGrid • Provide an online access points to TeraGrid • Aggregate domain science application-level capabilities • Hide the complexity of using TeraGrid • Web-based user environment is commonly used to meet such needs • Focus • Simplified user access through gateway user environment • Provide technological demonstrations for gateway development • Provide transparent access to TeraGrid for domain scientists • Build application-oriented workflow for easy access by users http://www.teragrid.org/gateways/

  10. SimpleGrid Gateway Design

  11. DMS Analysis Using Gateway User Environment • Client: click “New” to create a DMS job • Server: job submission • Invoking DMS Web service for job submission • Recording job in database • Client: automatic job status refresh through YUI timer • Server: handling job status refresh request • Invoking DMS Web service for job status • Updating database • Posting Twitter updates • Client: receiving Twitter updates periodically • Client: visualization • Drag&drop • Client-side JavaScript call for Google Map overlay creation and display

  12. DEMO

  13. Exercise • Conduct a DMS analysis • Name the analysis as “dms-yourname” • Submit an analysis • Follow SimpleGrid tweets • Refresh job status • Drag the finished job to Google map

  14. Demystifying the SimpleGrid Gateway Development • User environment development • Client side-centric user interactions • Contact server if necessary • Web 2.0 rich client that leverages JavaScript/Ajax libraries for Web content rendering, communications, interactions with external online services (e.g., Google Map and Twitter) • Web server acts as an information provider and communication proxy • TeraGrid functions • TeraGrid access is hidden from user environment • Integrated into backend service-oriented computing architecture • Grid-enable analysis application service development • Simplified programming interface • Streamlined service creation

  15. Tutorial Environment • Tutorial server • gw3.quarry.iu.teragrid.org • Tutorial server software • SSH client and server • Globus Toolkit 4.0.1 client tools • Java J2SE 1.6 • Ant 1.6.5 • Apache httpd 2.2 • Tomcat 5.5.27 • Axis2 1.4.1 • YUI 2.7.1 • PHP 5.1.6 • MySQL server 5.0.45 • TeraGrid sites • NCSA and SDSC

  16. User Accounts • TeraGrid training accounts • train21 – train50 • Gateway head nodes: • tg-login.ncsa.teragrid.org • tg-login.sdsc.teragrid.org • Tutorial server UNIX accounts • train01 – train20 • Password dispatched with tutorial materials • BASH shell environment • SimpleGrid location • $HOME/simplegrid2

  17. Part II: SimpleGrid Gateway User Environment Development Web client Visualization Collaboration Computation Messaging Map data Twitter Web Server Google Map Service invocation Web Services Grid-based computation TeraGrid

  18. Technologies • Web 2.0 development • JavaScript • AJAX • Yahoo UI (YUI) • Twitter API • Google Map • PHP-based server-side code development • Web service client programming • MySQL database programming

  19. Web 2.0 Technologies • Web 2.0 • A cloud • Portlets, AJAX, JSON, REST, Wiki, blogging, social networking, semantic Web, … … • Features • Collaborative Web content management • Dynamic and flexible Web content presentation • Focus • YUI-based Web 2.0 user interface development • Using Twitter to coordinate collaborative analysis

  20. User Interface: Rich Client vs Portal • Portal • Portal is basically server-side technology for Web content management • MVC (Model-View-Control), the common portal/portlet development model, happens at server side. • Portal server controls Web content rendering • Technology is well-developed • Web 2.0 rich-client user interface • Rich client • Highly flexible and dynamic control of Web content presentation through JavaScript and Web browser DOM • Server role change: Web server and Web services are used for data-feeding and message-handling. The control of Web content rendering happens at client • Technology is evolving

  21. User Interface: Rich Client vs Portal (Cont’d) • Impact on gateway user interface development • Portal and rich-client user interface are complementary to each other • Server side control: user management, request routing, access control, etc. • Client side user-friendliness. • Existing TeraGrid gateways are adding more and more Web 2.0 features into gateway portals. • SimpleGrid • SimpleGrid 0.7: Portal technology illustration • SimpleGrid 2.0: Web 2.0 user interface illustration

  22. Yahoo UI (YUI) • Web 2.0 user interface technologies • Yahoo UI (YUI) • 4-year history • Plenty of user interface components • Quick learning curve • Plenty of online code examples • jQuery • YUI features • User interface • Sliders, menus, layouts, client-server communications, tables,… … • Client-server communication • AJAX • Get utility for 3rd-party communication • Event and data handling • Drap and drop, timer, communications • JSON, plain text, XML data support • Development support • Dynamic JavaScipt library loading • JavaScript library compressor

  23. Outline for SimpleGrid Gateway Development • Developing Web 2.0 user environment • Layout design • Using YUI components to design a highly usable Web interface • Using Google Map for analysis result visualization • Using Twitter for collaboration • Web service invocation • Implementing application workflow in gateway • Illustrate a collaborative DMS analysis using the gateway

  24. SimpleGrid Gateway User Interface • Left pane: collaboration • Each gateway user participates a Twitter group for this tutorial • Web server is the group messaging coordinator • Notify computation progress through Twitter updates • Middle pane: visualization • Google Map-based visualization on DMS results • As results are available, users drag and drop a DMS job for visualization • Right pane: DMS computation management • Create and submit DMS jobs • Automatic job status refresh • Computation report

  25. Programming the SimpleGrid Gateway User Interface • Code location • $HOME/simplegrid2/webapp • YUI programming • YUI libraries and stylesheets • YUI data table programming • DMS job list retrieval, sorting, and rendering • DMS job selection, highlight, and drag&drop • YUI dialog programming for DMS job submission • Event handling for user actions and timers • Visualization • Invoking GISolve’s Google Map visualization library • Twitter programming • Retrieving Twitter updates on DMS job status change for all participants

  26. SimpleGrid Gateway Server • Apache+PHP+MySQL-based Web server • Apache for HTTP request handling • PHP for data feeding and Web service call invocation • MySQL for permanent job data storage • Functions • Serve Web content data • Index.html and other static Web resources (e.g., images) • JavaScript libraries • Handling DMS job creation, submission, refresh, listing, and deletion • Invoking DMS Web service through Php Web service client • Posting Twitter updates on job status change

  27. Programming the SimpleGrid Gateway Server • Code location • $HOME/simplegrid2/webapp/dms.php

  28. Exercise: Watch What happened Behind Gateway User Environment • Practise user controls • Submit one DMS job • View what happens behind at Web server, Web service, and TeraGrid • Drag&drop the submitted job into visualization pane • Watch other participants’ activities by following simplegrid posts on Twitter

  29. Part III: TeraGrid Services and Tools • Questions to answer • Where are your home directories on all 3 TeraGrid sites? • What is your grid certificate identity? • How do you create a grid proxy with a specified duration using MyProxy? • Are you able to transfer a sample dataset to a specified TeraGrid site? • How to submit a job and return immediately without waiting for its completion?

  30. Computation Flow

  31. Technologies • TeraGrid services • Software environment • CTSS (Coordinated TeraGrid Software and Services) • Resource management • Using GRAM/WS-GRAM to access TeraGrid HPC clusters • Data storage • Using GridFTP/RFT to access TeraGrid storage services • Information • Resource selection through TeraGrid information services • Security • Grid security • MyProxy

  32. Exercise: Setup Development Environment • Login to tutorial server • SSH login: username@${tutorial_server} • Environment setup cd ~ # setup grid environment . /opt/osg-client/setup.sh

  33. DMS Package Deployment on TeraGrid • Login to a TeraGrid cluster’s head node • SSH • TeraGrid user portal -> My TeraGrid -> SSH Terminal $ ssh train01@tg-login.ncsa.uiuc.edu • Check your software environment by looking at $HOME/.soft ## content of your .soft setting @teragrid-basic # TeraGrid wide Globus 4 and Grid software suite @globus-4.0 # Platform recommended development software suite @teragrid-dev • Check DMS package location $ ls $HOME/gisolve/DMS

  34. Get a Grid Proxy on Client $ myproxy-logon -l gisolve -t 100 -s myproxy.teragrid.org –p 7514 Enter MyProxy pass phrase: A credential has been received for user gisolve in /tmp/x509up_u502. $ grid-proxy-info subject : /C=US/O=National Center for Supercomputing Applications/CN=Gisolve Community User issuer : /C=US/O=National Center for Supercomputing Applications/CN=Certification Authority identity : /C=US/O=National Center for Supercomputing Applications/CN=Gisolve Community User type : end entity credential strength : 1024 bits path : /tmp/x509up_u502 timeleft : 99:59:54 (4.1 days) http://teragrid.org/userinfo/access/ http://myproxy.ncsa.uiuc.edu

  35. Transfer a Sample Dataset $ globus-url-copy file:$HOME/simplegrid2/webapp/datasets/sample gsiftp://gridftp-hg.ncsa.teragrid.org:2811/~/sample1 $ ssh gisolve@tg-login.ncsa.teragrid.org tg-login2 ac/gisolve> ls -l sample1 -rw-r--r-- 1 gisolve lpt 1116491 2009-05-29 21:28 sample1 http://teragrid.org/userinfo/data/transfer_location.php http://teragrid.org/userinfo/data/gridftp.php

  36. GRAM Job Submission • Compose job submission file in RSL format &(jobType="single") (count=1) (host_count="1") (project=TG-SES090019) (directory="/home/gisolve/gisolve/DMS/release") (executable=/home/gisolve/gisolve/DMS/release/process.pl) (arguments="500" "500" "20" "/home/gisolve/sample1" "/home/gisolve/result1") (stdout="stdout.gisolve.test") (stderr="stderr.gisolve.test") • Commands $ globusrun -b -f ./gt2.rsl -r grid-hg.ncsa.teragrid.org:2119/jobmanager-pbs globus_gram_client_callback_allow successful GRAM Job submission successful https://tg-grid1.uc.teragrid.org:50004/19686/1180493030/ $ globus-job-status https://tg-grid1.uc.teragrid.org:50004/19686/1180493030/ DONE

  37. WS-GRAM Job Submission • Compose job submission file in RSL XML format <job> <factoryEndpoint xmlns:gram="http://www.globus.org/namespaces/2004/10/gram/job" xmlns:wsa="http://schemas.xmlsoap.org/ws/2004/03/addressing"> <wsa:Address> https://grid-hg.ncsa.teragrid.org:8443/wsrf/services/ManagedJobFactoryService </wsa:Address> <wsa:ReferenceProperties> <gram:ResourceID>PBS</gram:ResourceID> </wsa:ReferenceProperties> </factoryEndpoint> <executable>/home/gisolve/gisolve/DMS/release/process.pl</executable> <directory>/home/gisolve/gisolve/DMS/release</directory> <argument>500</argument> <argument>500</argument> <argument>20</argument> <argument>/home/gisolve/sample1</argument> <argument>/home/gisolve/result1</argument> <stdout>/users/gisolve/gisolve/DMS/results/stdout.sample1</stdout> <stderr>/users/gisolve/gisolve/DMS/results/stderr.sample1</stderr> </job> $ globusrun-ws –b –submit –f my_rsl_xml http://teragrid.org/userinfo/jobs/globus.php

  38. TeraGrid Information Services • How do I know which TeraGrid site to use for my job submission and data transfers? • Check out TeraGrid web site for resource information • Use TeraGrid information services • Command: tginfo $ ./tginfo gram2 at ncsa HTTP Request to: http://info.teragrid.org/web-apps/csv/tginfo-v1/cli/gram2/at/ncsa Name ResourceID EndPoint --------------- ------------------------- ------------------------------------------------------- prews-gram-fork dtf.ncsa.teragrid.org grid-hg.ncsa.teragrid.org:2119/jobmanager-fork prews-gram-pbs abe.ncsa.teragrid.org https://grid-abe.ncsa.teragrid.org:2119/jobmanager-pbs prews-gram-pbs lincoln.ncsa.teragrid.org https://grid-abe.ncsa.teragrid.org:2119/jobmanager-pbs prews-gram-fork lincoln.ncsa.teragrid.org https://grid-abe.ncsa.teragrid.org:2119/jobmanager-fork prews-gram-pbs cobalt.ncsa.teragrid.org grid-co.ncsa.teragrid.org:2119/jobmanager-pbs prews-gram-fork cobalt.ncsa.teragrid.org grid-co.ncsa.teragrid.org:2119/jobmanager-fork prews-gram-fork abe.ncsa.teragrid.org https://grid-abe.ncsa.teragrid.org:2119/jobmanager-fork prews-gram-pbs dtf.ncsa.teragrid.org grid-hg.ncsa.teragrid.org:2119/jobmanager-pbs http://teragrid.org/userinfo/hardware/resources.php http://info.teragrid.org/

  39. TeraGrid-Based DMS Analysis Summary • Command line practice • Globus Toolkit 4.0, MyProxy • Typical computing process • Request an individual or community account on TeraGrid • Install DMS executables on three TeraGrid sites • Prepare a dataset on a local machine • Transfer a specified dataset to a TeraGrid site (e.g., NCSA)‏ • Submit a Grid job to the specified TeraGrid site with a parameter value • The submitted job is scheduled to be executed on one compute node on the specified TeraGrid cluster • When the job is finished, the analysis result is written into the data directory of DMS installation on the TeraGrid cluster • Transfer the result back to the local machine • Visualize the result using the DMS visualization tool • Learning experience • TeraGrid resource access using Globus Toolkit • Conduct a scientific computing process using command line tools • Realizing that manual Grid access is error-prone and not scalable

  40. Part IV: SimpleGrid API and Web Service Development for Grid-enabled Analysis • How to do Part II in a programming way? • Purpose • Streamline the access to TeraGrid resources as a Grid-enabled application • Build Application Web services for interactions with gateway user environment • Package • Location: $HOME/simplegrid2/java • SimpleGrid API for DMS analysis • Libraries • All jars from ${GLOBUS_LOCATION}/lib • Particularly cog-jglobus.jar • Apache Axis2 http://www.globus.org/cog/manual-user.pdf

  41. Programming Techniques • Java CoG kit API • Grid security programming • MyProxy programming • SAML attribute-based gateway user authentication programming • GRAM and WS-GRAM programming • GridFTP programming

  42. SimpleGrid API • SimpleCred: Grid credential management • Proxy creation: local and MyProxy • Automatic proxy renewal • Attribute-based X.509 credential for authorization • SimpleTran: Data transfer to/from grids • GridFTP • SimpleRun: Grid job management • GRAM and WS-GRAM • Batch job submission • SimpleViz: Visualization component • Java 2D + Google map (shown later) • SimpleInfo: Grid information provider * * Under development

  43. SimpleGrid API architecture • Learning experience • Java-based Grid-enabled scientific application development • Application deployment is a constraint for large-scale user community

  44. Get a Grid Proxy • Class • org.simplegrid.grid.security.SimpleCred • Two methods to get a proxy • Load an existing valid proxy file • SimpleCred::load() • Create a proxy by contacting with a MyProxy server • SimpleCred::logon() • Generalization: SimpleGrid::get() • Export proxy to a file • SimpleCred::export()

  45. Transfer a Sample Dataset • Class • org.simplegrid.grid.data.SimpleTransfer • Local<->remote transfer • SimpleTransfer::remote2local() • SimpleTransfer::local2remote() • Third-party transfer • Refer to Cog manual (version 1.1) http://wiki.cogkit.org

  46. GRAM Job Submission • Class • org.simplegrid.grid.job.SimpleRunGT2 • org.simplegrid.grid.job.SimpleRSL • RSL generation • Application specific • GT2 job submission • Use batch mode in which a call returns immediately with a job handle

  47. GRAM Job Submission: MPI • MPI on TeraGrid • User “softenv | grep mpi” to see different MPI settings • MPI job submission through Globus • Specify “count” element in RSL • “count” is the number of processes to spawn • Specify “host_count” element in RSL • “host_count” is the number of CPUs for MPI execution • “count = host_count” means each CPU runs one MPI process • Specify “jobType=mpi” in RSL

  48. WS-GRAM Job Submission • Class • org.simplegrid.grid.job.SimpleRunGT4 • org.simplegrid.grid.job.SimpleRSL • RSL XML generation • Application specific • GT4 job submission • Use batch mode in which a call returns immediately with a job handle • Note that GT4 uses a different globus package org.globus.exec.* http://www.globus.org/toolkit/docs/4.0/execution/wsgram/developer-index.html#id2563059

  49. How to Write RSL? • RSL schema • http://globus.org/toolkit/docs/4.0/execution/wsgram/schemas/gram_job_description.html • GT2 and GT4 schema comparison • http://www.globus.org/toolkit/docs/4.0/execution/wsgram/WS_GRAM_Migrating_Guide.html

  50. Exercise 2 • Test SimpleGrid API • Source simplegrid-env.sh • Compile: $HOME/simplegrid2/bin/AppCompile.sh • Run: $HOME/simplegrid2/bin/AppRun.sh

More Related